Biomedical Engineering Reference
In-Depth Information
Fig. 5.2
Example of a knee joint multiscale dataset: (
a
) Cross-section of knee CT scan, (
b
)micro-
CT slice of cartilage tissue, (
c
) 3D reconstruction of micro-CT scan of cartilage tissue, (
d
) histo-
logical image of cartilage, (
e
) schematic of extracellular components of cartilage tissue, (
f
) JMol
visualization of aggrecan particle [
7
]. Images (
b
,
d
) and dataset (
c
) courtesy of 3B's Research
Group. Dataset (
a
) courtesy of OsiriX Project [
8
]
High-dimensional image data:
Images carry more information in the form of
additional dimensions (beside x- and y-resolution): e.g. time, space and channels.
For instance, multispectral imaging (acquisition of spectrally resolved information
at each pixel of an imaged scene) has become widely offered by microscopy
manufacturers [
2
].
Amount of data:
Massive datasets from large scale experiments are difficult to
manage due to the memory limitations. Even though the amount of available
memory is increasing and out-of-core techniques have been developed [
3
], data
memory requirements increase due to a more detailed data collection. This creates
another challenge: representing datasets in a user intuitive manner is more difficult
in relation to the increasing amount of data.
3D reconstruction and projection techniques are important when dealing with
high-dimensional image data. This is illustrated in an example of tomographic min-
eralogical data analysis [
4
]. The software used (YaDiV [
5
]) allowed the experts from
mineralogy to understand the geometric spatial structure intuitively, which was not
observed in the respective 2D slice images used before. This investigationmethod can
also be applied in the context of biomedical multiscale visualization, where recon-
struction of micro-CT data of cartilage clearly exhibits complicated spatial tissue
structures (Fig.
5.2
c).
The
multimodal requirement
is born by this variety of data properties: multiple
imaging sources provide vast amounts of data with heterogeneous dimensionality
that should be merged. This requirement is needed to help physicians and scientists
of the same domain to interpret this wide range of collected data [
6
].
